1. Field of the Invention
The present invention relates to an automatic focusing device for a film scanner which picks up image signal from an image photographed on negative or positive film.
2. Background Arts
The film scanner is used for picking up image signal from photographic image in order to simulate the photographic image as a video image on a monitor, or to make a hard copy of the photographic image through a video printer. The film scanner has an image sensor or a solid state imaging device which scans the film surface to photoelectrically convert the photographic image into electric signal. The electric signal is processed into appropriate image signal.
An imaging lens is disposed in between the film surface and the image sensor, so that the photographic image is focused on a photoelectric conversion surface of the image sensor. In the film scanner, the filmstrip is held in a film carrier which is placed at a substantially constant distance from the imaging lens. Therefore, the imaging lens of the film scanner needs not cover such a wide focusing range as necessary for ordinary video cameras and electric still cameras. But it is still necessary to adjust the focus of the imaging lens as the distance to the film surface can vary depending upon the type of filmstrip as having different thickness, or when the film carrier is interchanged.
Automatic focusing devices have been known in the art. For example, JPA 61-41277, JPA 63-215268 and JPA 1-7774 disclose an automatic focusing device which basically scan the photographic image and samples changes or differences in density of the image between sampling positions, and produces evaluation signal which correlates the sampled differences with their sampling number. Then, the evaluation signal is integrated. While changing the focal point of the imaging lens, the automatic focusing device detects integration value of the evaluation signal at regular intervals, and checks increment or decrement of the integration value to find out a peak of the integration value. The imaging lens is set at a position where the integration value is at the peak. This is because the peak of the integration value of the evaluation signal indicates the in-focus position of the imaging lens.
FIG. 13 shows magnitude distribution curves of the evaluation signal with respect to spatial frequency, wherein the spatial frequency corresponds to the sampled difference or the change between the sampled densities, while the magnitude of evaluation signal corresponds to the sampling number. The three curves are obtained in three different focusing positions: the curve shown by a solid line represents evaluation signal obtained at an out-of-focus position, the curve shown by dashed lines represents evaluation signal at an in-focus position, and the curve shown by chain-dotted lines represents evaluation signal obtained on the way to the in-focus position. The areas bounded by the individual curves correspond to the integration values of the respective evaluation signals. As the imaging lens is being focused on the photographic image, the more evaluation signal is obtained from the higher spatial frequency range, so the integration value of the evaluation signal increases. Accordingly, it is possible to consider that the in-focus position of the imaging lens is where the integration value of the evaluation signal is at the peak. Hereinafter, changing the focusing position for sampling evaluation signal will be referred to as “focus scanning”.
FIG. 14 shows relationship between the focusing position of the imaging lens and the integration value of the evaluation signal. A curve C1 represents an ideal curve wherein the peak and thus the in-focus position is definite and easy to determine. In order to obtain such an ideal curve, it is necessary to integrate those evaluation signals which belong to an appropriate spatial frequency band. That is, if a frequency band B1 is optimum for a photographic image, but a too narrow frequency band B2 is selected, as shown in FIG. 15, the curve of the consequent integration value would be as shown by C2. As the curve C2 waves around an actual peak, a subsidiary peak is apt to be mistaken as the peak. If a too wide frequency band B3 is selected in that case, the curve of the consequent integration value would be as shown by C3. The peak of the curve C3 is indefinite, and thus the in-focus position is indefinite.
In general, when the photographic image is low contrast, integration value of the evaluation signal changes gently along with the focusing, as shown by the curve C3. On the contrary, where the density of photographic images varies frequently over a wide range, integration value of the evaluation signal changes sharply and unstably in a range around the in-focus position, as shown by the curve C2.
Since the above mentioned prior arts use a band pass filter which selects a predetermined frequency band of the evaluation signal for any kinds of photographic images, it is sometimes difficult to accurately determine the in-focus position. Accordingly, in order to improve the focusing accuracy, it is desirable to select an appropriate frequency band of the evaluation signal in accordance with the photographic image. However, as there are a variety of image patterns, it is difficult to automatically select an optimum frequency band for every photographic image.
In addition, since the imaging lens of the film scanner is often a zoom lens so as to permit scanning a limited portion of the photographic image within a frame, and the spatial frequency components of the signal from the image sensor change with the zooming, it is also necessary for accuracy to adjust the frequency band to the change in spatial frequency of the image signal.
To avoid mistaking the peak when the integration value of the evaluation signal has such a curve as shown by C2 in FIG. 14, the prior arts teaches to move the imaging lens through the entire focusing range while sampling integration values, thereafter determine a peak value of these integration values, and set the imaging lens to the position where the peak is obtained. However, this method is so time consuming that it does not work in practice.